This study compared seven publically available and commonly used de novo assembly tools: SSAKE, VCAKE, Euler-sr, Edena, Velvet, ABySS and SOAPdenovo. These tools are specifically designed to assemble large numbers of short reads generated by next-generation sequencing platforms.
In analyzing these tools, stronger performance is indicated by higher N50 values, higher sequence coverage, lower assembly error rates and lower computational resource consumption (to enable assembly of larger genomes). The performance of different assembly tools was dependent, to some extent, on the test conditions. Based on the results of our investigation, we propose the following guidelines for tool selection. Generally, SSAKE, Edena and Euler-sr need higher depths of coverage (~50×) than Velvet, ABySS and SOAPdenovo (~30×) to generate higher N50 lengths; SOAPdenovo was the fastest of all tools, and ABySS almost always consumed the least memory space. We have developed a tentative reference/guidelines for selecting optimal de novo tools under varying conditions (). Specific comments regarding the performance of individual tools under different conditions are summarized below.
Recommendations for de novo tool selection under varying conditions
SSAKE provided good sequence coverage, and also generated good N50 lengths when assembling sequences with low GC content. On the other hand, SSAKE tended to generate more assembly errors and needed more depth of coverage to reach DCAP than most of the other tools tested. The time and memory usage of SSAKE was also the highest of the tools tested. Our results indicated that assembly of large sequences (e.g. Homo sapiens) was not feasible with SSAKE.
VCAKE produced the shortest N50 lengths in most situations, and the sequence coverage by VCAKE was comparable to SSAKE. VCAKE also generated many assembly errors, even higher than that of SSAKE under certain test conditions. The computational resources required to run VCAKE were a little less than those required for SSAKE.
In assembling single-end short reads, Euler-sr produced the longest N50 values, but it also generated high assembly error rates, comparable to that of SSAKE. In addition, sequence coverage of Euler-sr was the lowest under most test situations. Euler-sr consumed intermediate computational resources.
Under most conditions tested, Velvet and ABySS show similar assembly performance; they generated similar N50 lengths, their DCAPs were relatively low and they required acceptable computational resources. Consequently, it is feasible to use these tools for assembling large sequences, such as those obtained for Homo sapiens. ABySS produced fewer assembly errors, and consumed a little less memory and more runtime than Velvet. When assembling paired-end reads, ABySS produced the highest assembly coverage of all tools tested. When assembling larger genomes, Velvet sometimes used exceptionally high runtimes and memory.
Edena needs a high depth of coverage, comparable to SSAKE, to reach the DCAP. It produced similar, or greater, N50 values to Velvet in most single-end assemblies, and generated assembly error rates that were comparable to Velvet. The computation demands of Edena were intermediate, between SSAKE and ABySS.
SOAPdenovo was the fastest assembler. Its DCAP was as low as that of ABySS and it produced among the highest N50 values in paired-end read assembly, and relatively high N50 values in single-end assembly. SOAPdenovo generated higher assembly error rates and lower sequence coverage than ABySS. It also consumed more memory than ABySS when assembling paired-end reads. The appropriate setting for SOAPdenovo (SOAPdenovo31mer, SOAPdenovo63mer and SOAPdenovo127mer that support kmer ≤31,≤63 and ≤127, respectively) must be selected based on read length. SOAPdenovo63mer/SOAPdenovo127mer consumed two/four times as much RAM as SOAPdenovo31mer.
In light of our results, investigators may choose the most appropriate assembly tool(s) to use based on their specific experimental setting and available computational resources. Our results may also serve as a reference, when designing sequencing projects, for selecting targeted depths of coverage, control levels of sequencing error rates, etc. Given the rapid increase in use of next-generation sequencing technologies, our results should be of value to both empiricists, during experimental design, and to bio-informaticians who seek guidance for selecting appropriate assembly tool(s) for data analyses and who attempt improvement of the assembly tools.
Funding: Shanghai Leading Academic Discipline Project (S30501 in part); startup fund from Shanghai University of Science and Technology. The investigators of this work were partially supported by grants from NIH (P50AR055081, R01AG026564, R01AR050496, RC2DE020756, R01AR057049 and R03TW008221); Franklin D. Dickson/Missouri Endowment from University of Missouri–Kansas City and the Edward G. Schlieder Endowment from Tulane University.
Conflict of Interest: none declared.